Professional Documents
Culture Documents
Abstract
This paper reports the results of an analytical, numerical and experimental investigation
dealing with I-section steel members acted by a combination of major-axis bending and axial
(“beams subjected to tension”), which is relatively rare in practice and, therefore, has received
little attention in the past. In particular, there are no guidelines for the design against buckling
ultimate limit states of such members (only their cross-section resistance is checked). This
means that the axial tension favorable effect on lateral-torsional buckling/failure is neglected,
leading to over-conservative designs − indeed, a beam subjected to axial tension is
currently designed against lateral-torsional failure as a “pure beam”. In order acquire scientific
knowledge and provide design guidance on this topic, the lateral-torsional stability, failure and
design of hot-rolled steel I-beams with fork-type end supports and acted by simple transverse
loadings (mostly applied end moments) and various axial tension values are addressed in this
work. After developing and validating an analytical expression to calculate critical buckling
moments of beams under uniform bending and axial tension, beam finite element buckling
results are presented for the non-uniform bending cases. Then, two full-scale tests involving
a narrow and a wide flange beam under eccentric tension are described and their results are
used to develop finite element models, subsequently employed to perform a parametric study
aimed at gathering a fairly extensive ultimate strength/moment data bank. Finally, this data
bank is used to assess the merits of a design approach proposed here for beams subjected to
tension and collapsing in lateral-torsional modes − this design approach, which consists of
slightly modifying the current procedure prescribed in Eurocode 3 to design beams against
lateral-torsional failures, is shown to provide ultimate moment estimates that correlate very
well with the values obtained from the numerical simulations.
1. Introduction
In recent years, the technical and scientific community dealing with steel structures has
devoted a considerable effort to the development of efficient (safe and economic) procedures
and formulae (interaction equations) for the design and safety checking of steel members
(i) subjected to different combinations of internal forces and moments and (ii) susceptible to
global instability phenomena, namely flexural buckling (members under compression) and/or
lateral-torsional buckling (open-section members under major-axis bending). Moreover, the
vast majority of the existing studies concern I-section members, by far the most widely used
in the steel construction industry. This fact is reflected in the very large number of “fine-
tuned” expressions (interaction equations), intended for the design and safety checking of I-
section members, appearing in the current steel design codes. For instance, the latest version of
1
Civil Engineering Department, ICIST, Instituto Superior Técnico, Technical University of Lisbon, Portugal.
2
Civil Engineering Department, University of Applied Sciences of Western Switzerland (Fribourg), Switzerland.
669
Eurocode 3 (CEN 2005) contains a plethora of rather elaborate (and also fairly complex)
formulae and equations aimed at the design of narrow-flange (I) and wide-flange (H) cross-
sections and members subjected to arbitrary internal forces and moment diagrams − the
interested reader can find the background of most of these formulae and equations in the
ECCS (European Convention for Constructional Steelwork) report stemming from the activity
of its Technical Committee on Stability (TC8) and co-authored by Boissonnade et al. (2006).
However, quite surprisingly (in view of what was mentioned in the previous paragraph),
virtually no information can currently be found concerning the structural response and design
of I-section members subjected to major-axis bending and tension (i.e., beams under tension),
namely on how the presence of tension affects (improves) the beam lateral-torsional buckling
behavior. Indeed, the rather complete literature search carried out by the authors bore no fruits
and, moreover, no information was obtained from several world-wide recognized experts on
lateral-torsional buckling that were contacted very recently. The sole exception to the above
situation is the previous (pre-norm) version of Eurocode 3 (CEN 1992), which included
provisions concerning the safety checking of beams under tension. Such provisions, whose
existence provided the motivation for the study reported in this work, are based on an
“effective (reduced) bending moment” concept − however, once more, no trace of background
information concerning these provisions could be found. Of course, part of the explanation
for the “information void” on this problem is due to the fact that (i) beams under tension
occur seldom in practice and (ii) neglecting the tension effects leads to conservative ultimate
strength estimates against lateral-torsional failures (the member is designed or safety checked
as a “pure beam”). Nevertheless, it is important to investigate the behavior, collapse and
design of beams susceptible to lateral-torsional buckling and subjected to tension, namely to
acquire information on how conservative are the ultimate strength predictions that neglect the
tension effects. The objective of this paper is precisely to contribute to such an investigation,
by bridging the lack of scientific information and technical guidance concerning the lateral-
torsional stability, behavior/failure and design of beams under tension. It deals specifically
with (doubly symmetric) hot-rolled steel I-section beams exhibiting “fork-type” end supports
and subjected to simple transverse loadings (e.g., applied end moments) and not affected by
local buckling phenomena − beams with a compact cross-section (class 1 or 2 cross-section,
according to the Eurocode 3 nomenclature) that can reach its plastic resistance.
Initially, the paper presents the derivation and validation, through the comparison with beam
finite element results, of an analytical expression that provides critical buckling moments
associated with the lateral-torsional stability (bifurcation) of uniformly bent beams subjected
to tension. Then, the analytical study is (numerically) extended to beams under non-uniform
bending (mostly stemming from unequal applied end moments) − several beam finite element
results concerning the beneficial influence of axial tension on the lateral-torsional stability
are presented and discussed in some detail. Next, the paper describes the experimental set-
up, procedure and obtained results concerning two full-scale tests, involving a narrow and
a wide flange beam subjected to eccentric axial tension, that were performed (i) to acquire in-
depth knowledge about the problem under consideration and (ii) to provide the means to
develop and validate beam and shell finite element models, in the software FINELG (2012), to
subsequently perform an extensive parametric study. This parametric study, carried out in
order to assemble a fairly large ultimate strength/moment data bank, involves more than 2000
numerical simulations concerning beams with various cross-section shapes, lengths, yield
stresses, acting bending moment diagrams and axial tension levels. This data bank is then
used to assess the merits of a design approach proposed in this work for beams subjected to
tension and collapsing in lateral-torsional modes − this design approach, which consists of a
670
slight modification of the current procedure prescribed in Eurocode 3 to design beams against
lateral-torsional failures, is shown to provide ultimate moment estimates that correlate very
well with the values obtained from the FINELG numerical simulations.
(a) (b)
Figure 1: Beam subjected to major-axis bending My and axial tension Nt: (a) general view and (b)
deformed configuration associated with the occurrence of lateral torsional buckling
Following the derivations presented in the classical monographs by Chen & Atsuta (1977)
and Trahair (1993) for the lateral-torsional stability of beam-columns (i.e., members under
major-axis bending and axial compression), the equations ensuring adjacent equilibrium for
members subjected to major-axis bending and axial tension read (note the change in sign of
the Nt terms and recall that pre-buckling deformations are neglected)
where v and φ are the minor-axis bending displacements and torsional rotations, respectively.
In order to obtain the critical buckling moment, it is necessary to solve the eigenfunction
problem defined by Eqs. (1) and (2). Following the approach adopted by Culver (1966) to
obtain analytical solution for beam-columns subjected to bending, it can be shown that the
beam buckling mode is expressed by the sinusoidal functions
671
v(x) = A1 sin (π/L x) (3)
satisfying all the problem boundary conditions, it is possible to obtain the expression
(5)
which provides the critical buckling moment of a member subjected to bending and tension.
In this expression, (i) Mcr (0) denotes the critical buckling moment of the “pure beam”
(member under uniform bending only), and (ii) Pcr.z and Pcr.φ are given by
(6)
(7)
and their values correspond to the symmetric of the minor-axis flexural and torsion buckling
loads of the “pure column” (member under uniform compression only). Eq. (5) confirms
(and quantifies, for the particular case under consideration) the beneficial effect of tension
on the member lateral-torsional buckling moment − i.e., the additional bending and torsional
stiffness values, stemming from the presence of Nt, lead to a Mcr increase.
In order to provide validation for the derived analytical expression, Fig. 2 depicts the
variation of the critical bucking moment increase [Mcr (Nt) /Mcr (0)] with the ratio Nt /My, for an
IPE 300 beam with length L=10 m − the results shown are obtained with Eq. (5) (solid line)
and ABAQUS BFE buckling analyses (dashed line). It is observed that (i) there is a virtually
perfect agreement between the analytical a numerical values, and (ii) the critical moment
increase grows exponentially with applied tension level − for Nt /My larger than 9.6, lateral-
torsional buckling no longer occurs (the whole beam is under tension).
Figure 2: Variation of the critical buckling moment increase Mcr (Nt) /Mcr (0) with Nt (IPE 300 + L=10 m)
672
2.2 Beams under Non-Uniform Bending − Numerical Investigation
ABAQUS BFEA analyses are employed to determine the critical buckling moments of beams
(i) built from four hot-rolled profiles (IPE300, IPE500, HEB300, HEB500), (ii) with lengths
comprised between 0.5 and 25 m, (iii) subjected to the bending moment diagrams stemming
from the loadings depicted in Fig. 3 (unequal end moments or uniformly distributed load
applied along the centroid/shear centre axis) and (iv) acted by tension values such
thatβ=Nt /My=0; 0.5; 0.75; 1.0; 1.5; 2.0 (My is the maximum applied bending moment) −
the transverse and axial loads are applied proportionally.
Due to space limitations, only a representative sample of the numerical critical buckling
moments obtained is presented in this work − the interested reader can find the full set of
results (qualitatively similar to those displayed here) in Tomás (2013). Fig. 4 concerns
IPE300 beams subjected to the bending moment diagram associated with ψ=0 and provides
the variations of the critical moment increase [Mcr (Nt) /Mcr (0)] with the ratio β=Nt /My
(relating the applied tension and maximum bending moment values) for span lengths varying
from L=0.5 m to L=15 m. On the other hand, Fig. 5 concerns HEB300 beams with length
L=10 m and provides curves the Mcr (Nt) /Mcr (0) vs. β for all the bending moment diagrams
considered in this work. Finally, Fig. 6 concerns L=15 m beams with various cross-sections
Figure 4: Variation of Mcr (Nt) /Mcr (0) with β for 0.5 m ≤ L ≤ 15 m (IPE300 beams + ψ =0)
673
Figure 5: Variation of Mcr (Nt) /Mcr (0) with β for various moment diagrams (HEB300 beams + L=10 m)
Figure 6: Variation of Mcr (Nt) /Mcr (0) with β for different cross sections (L=15m + ψ=0)
acted by one end moment (ψ=0) and shows additional Mcr (Nt) /Mcr (0) vs. β curves. The
observation of the results presented in these three figures, as well as those not shown here
but reported by Tomás (2013), prompts the following comments:
(i) Naturally, the critical moment increase always grows with β, i.e., the amount of tension
acting on the beam. Note that, eventually, the Mcr (Nt) /Mcr (0) vs. β curves would tend
to infinity as β approaches the “limit tension value” − the value ensuring that no beam
cross-section is acted by compressive stresses. Note, however, that a Nt value established
a priori, corresponding to the application of transverse loading on a “pre-tensioned”
beam, never precludes the occurrence of elastic lateral-torsional buckling (for any Nt).
674
(ii) The three sets of Mcr (Nt) /Mcr (0) vs. β curves, corresponding to the variation of either the
beam length (Fig. 4), bending moment diagram (Fig. 5) or cross-section shape (Fig. 6),
exhibit qualitatively similar trends, which means that it can be argued that the three
above factors are equally relevant.
(iii) Fig. 4 shows that, logically, the importance of the tension effects grows with the beam
length, i.e., as the beam becomes more prone to lateral-torsional instability.
(iv) Fig. 5 makes it possible to assess how the (beneficial) tension effects vary with the
bending moment diagram shape3. The key factor is the shape of the axial force diagram
acting on the whole beam compression flange, combining the constant tensile value Nt
with varying compressive values due to the bending moments. It is observed that this
combination leads to tension effects that are (iv1) highest for the ψ=0 triangular diagram
(“least compressed” compression flange) (iv2) lowest for theψ=1 uniform diagram
(“most compressed” compressed flange). It is interesting to note that the curve associated
with the uniformly distributed load falls, somewhat surprisingly, in between those
concerning the ψ= − 0.5 and ψ=0.5 diagrams.
(v) Finally, Fig. 6 provides clear evidence that the relevance of the tension effects grows
considerably with the web height. On the other hand, no conclusion can be drawn about
the influence of the flange width, which seems to depend also on the web height.
It is worth noting that the numerical investigation just reported made it possible to gather a
critical buckling moment data bank comprising about 1000 values. They will be used later, in
section 5, addressing the development of a design procedure for beams subjected to tension.
3
Note that the curves provide critical buckling moment percentage increases (not the values).
675
(a) (b)
Figure 7: Experimental set-up (a) overall view and (b) detail of the end supports
Figure 8: Procedure associated with the application of the eccentric tensile load
Prior to the performance of each test, (i) tensile coupon tests were carried out, in order to
assess the real steel material behavior, (ii) the residual stress distribution and magnitude were
quantified, by dividing (cutting) the profile into several strips and measuring their individual
elongations, and (iii) the specimen initial geometrical imperfections (global and local)
were assessed − this was done my measuring, at 10 cm intervals along the specimen length,
vertical displacements at three upper flange points (web-flange corner and flange free ends)
and lateral displacements at three web points (mid-height and web-flange corners).
When performing a test, the first steps consisted of (i) welding vertical rigid profiles to the
specimen ends (see Fig. 7(a)), which prevent warping and make it possible to apply the
eccentric tensile loads, (ii) positioning the specimen in between two pairs of end support
cylindrical hinges, one resting on the supporting cross-bar and the other leaning vertically
against a short RHS cantilever (see Fig. 7(b)4, making sure that symmetry with respect to the
mid cross-section is kept, i.e., that the outstand segments extending beyond each hinge are
equal (see Fig. 7(a), and (iii) placing the hydraulic jacks, which are mounted on secondary
structural systems, in such a way that the required tension eccentricity is guaranteed (see
Fig. 8). The second steps concern the placement of the measuring devices at the mid-span and
end cross-sections. In view of the expected specimen three-dimensional behavior, a complex
displacement transducer system was devised to enable the measurement of two pairs of mid-
span cross-section transverse displacements (two vertical and two lateral) − Fig. 9(a) makes it
possible to visualize this displacement transducer system. Moreover, inclinometers were
attached to the vertical rigid profiles welded to the specimen ends, in order to measure the
major-axis flexural rotations on the supports, as illustrated in Fig. 9(b). During the test, the
above measurement devices recorded values at 0.5 second intervals, thus providing a fairly
4
This support arrangement ensures free major/minor flexural rotations. Moreover, two pairs of short (rigid) steel
plates were placed laterally near each support (see Fig. 7(b)), to limit the end section minor flexural rotations.
676
(a) (b)
Figure 9: (a) Mid-span displacement sensor system at and (b) inclinometer measuring support flexural rotations
The results obtained during the performance of each test consist of (i) experimental evidence
concerning the occurrence of lateral-torsional buckling in beams acted by tension (Figs.
677
(a) (b)
Figure 11: IPE200 specimen (a) longitudinal normal residual stress distribution and (c) initial vertical
displacement longitudinal profiles of the three top flange points
(a) (b)
Figure 12: Deformed configurations at the brink of collapse for specimens (a) IPE200 and HEA160
(a) (b)
Figure 12: Specimen IPE200 equilibrium paths relating the applied load to (a) the end section flexural
rotations θy and (b) mid-span torsional rotation θx
The results obtained during the performance of each test consist of (i) experimental evidence
concerning the occurrence of lateral-torsional buckling in beams acted by tension (Figs.
12(a)-(b) display the deformed configurations of the IPE200 and HEA160 specimens at the
brink of collapse), (ii) equilibrium paths relating the applied load with various measured
displacements and/or rotations (Figs. 13(a)-(b) show the IPE200 specimen equilibrium paths
678
concerning the end cross-section flexural rotation θy and mid-span torsional rotation θx) and
(iii) experimental failure loads: 269.2 kN (IPE200) and 150.0 kN (HEA160).
Figure 13: Overall view of the experimental set-up discretization concerning the IPE200 specimen
(a) (b)
Figure 14: (a) Initial geometrical imperfections (amplified) and (b) load application system models (IPE200)
The developed numerical model was then employed to perform elastic buckling and elastic-
plastic geometrically non-linear numerical analyses of the two specimens tested. Table 1
makes it possible to compare the numerical and experimental ultimate moments Mu obtained
− moreover, the table also provides the (i) analytical cross-section plastic moments Mpl (under
pure bending), (ii) FINELG beam critical moments Mcr (analyzed in the experimental set-up
679
context − Fig. 15 displays half of the critical (lateral-torsional) buckling mode shape
obtained for the IPE200 specimen) and (iii) beam slenderness values, calculated on the basis
of the presented Mpl and Mcr values − note that both beam are fairly stocky.
Table 1: Analytical, numerical and experimental results concerning the two specimens tested
Numerical Experimental
Mpl [kNm] Mcr [kNm] Mu [kNm] Mu [kNm]
IPE 200 70.6 0.90 86.5 62.98 67.3
HEA 160 75.3 0.66 179 76.9 75
Figure 15: Half of the critical (lateral-torsional) buckling mode provided by FINELG for the IPE200 specimen
The numerical and experimental ultimate moments are found to correlate quite well − indeed,
the numerical simulations either underestimate by 6% (IPE200 specimen) or overestimate by
2% (HEA160 specimen) the experimental values. Moreover, Fig. 16 shows the IPE200
specimen deformed configuration at collapse, provided by the FINELG analysis − note the
qualitative and quantitative similarity with its experimental counterpart, shown in Fig. 12(a).
Figure 16: Deformed configuration at collapse obtained with FINELG for the IPE200 specimen
Finally, Figs. 17(a)-(b) compare the experimental equilibrium paths shown in Figs. 12(a)-(b),
relating the applied load to the end section flexural rotations and mid-span torsional rotation,
with the corresponding numerical simulations. First of all, note that, as mentioned earlier, the
numerical and experimental ultimate loads/moments practically coincide (2% difference).
Concerning the end section flexural rotations and mid-span torsional rotation, there is a
680
(a) (b)
Figure 17: Experimental and numerical equilibrium paths concerning the IPE200 specimen and relating
the applied load with the (a) end-section flexural rotations θy and (b) mid-span torsional rotation θx
virtually perfect coincidence in the elastic regime, beyond which the numerical model is a bit
stiffer and, therefore, underestimates the θy and θx values. A possible explanation for the
discrepancies observed between the numerical and experimental equilibrium paths resides
in the fact that the three-dimensional nature of the beam deformed configurations may be the
source of erroneous measurements. Indeed, it was observed that the accuracy of the transducer
measurements decreased considerably when the point under consideration exhibited various
displacement components. Although some corrections were made, on the basis of geometry
considerations, they gradually become less as the beam deformation increases, rendering
almost inevitable the underestimation of the measured displacements and rotations.
On the basis of the comparisons between the test results obtained and the corresponding
numerical simulations, most of which were not presented here and can be found in Tomás
(2013), it seems fair to conclude that the numerical (shell finite element) model developed
provides reasonably accurate results and, thus, can be adequately used to validate the finite
element beam model employed perform the parametric study addressed in the next section.
681
Figure 18: Constitutive law adopted to model the steel material behavior (Boissonnade & Somja 2012)
(a) (b)
Figure 19: (a) Residual stresses and (b) global initial geometrical imperfections considered in the numerical
analyses performed in the context of the parametric study (Boissonnade & Somja 2012)
FINELG and the beams were discretized into standard 3D beam finite elements based on
Vlasov’s theory for open-section thin-walled members5.
Due to space limitations, only a small (but representative) fraction of the results obtained
from the above parametric study can be shown in this paper, namely in Figs. 20-22 and
Annex A (tabular form − Tables A1-A4) − they concern the results obtained for (i) L=15 m
S355 IPE300, (ii) L=5 m S460 IPE500, (iii) L=10 m S355 HEB300 and (iv) L=3.5 m S355
HEB500 beams, all subjected to the various bending moment diagrams. Again, a detailed
report of the whole set of determined results can be found in Tomás (2013). The results
displayed in Figs. 20 and 21 concern the influence of the axial tension level on the ultimate
strength of IPE300 and HEB300 beams made of S460 steel, exhibiting various lengths,
comprised between L=0.5 m and L=15 m, and subjected to several bending moment
diagrams, all stemming from applied end moments. Both figures provide the variation of the
ultimate moment Mu, normalized with respect to the cross-section plastic bending resistance
Mpl (calculated for pure bending on the basis of fy), with the loading ratio β=Nt /My − the
values between parentheses, given above or below each point (beam analyzed) provide the Mu
percentage increase due to axial tension: [Mu (β) − Mu (0)] /Mu (0). While Fig. 20 focuses on the
combined effect of β and the beam length (under uniform bending), Fig. 21 addresses the
5
For validation purposes, the results obtained from FINELG beam and shell finite element analyses were compared
− the latter were based on the model validated against the experimental results (see section 3). Provided that
the beam length is long enough to preclude local buckling effects, a virtually perfect match was found.
682
Figure 20: Variation of Mu/Mpl with β and the beam length (S460 steel IPE300 beams under uniform bending)
Figure 21: Variation of Mu/Mpl with β and the bending moment diagram (L=15 m S355 steel HEB300 beams)
joint influence of β and the bending moment diagram (for a L=15 m beam). It is worth noting
that the negative and underlined positive values in Figs. 20 and 21 correspond to beams
whose collapse is governed by the cross-section resistance, which naturally decreases with β6.
The observation of the numerical results displayed in these figures, as well as those not shown
here but reported by Tomás (2013), makes it possible to draw the following conclusions:
(i) First of all, as expected, the presence of axial tension is completely different in the
stocky and slender beams, in the sense that their collapse is governed by plasticity and
6
The Mu values provided in Tables A1-A4 are those that effectively correspond to the limit points obtained from
the FINELG non-linear analyses, which take into account the small strain-hardening depicted in Fig. 18.
However, the Mu values displayed in Fig. 20 are limited by the cross-section resistance determined analytically,
which is given in Tables A1-A4 (see the MN,Rk values) and neglects any strain hardening.
683
instability effects, respectively. In the former (e.g., the L=0.5; 1.0 m beams in Fig. 20),
axial tension leads to an ultimate moment decrease, stemming exclusively from the
drop in cross-section resistance. In the latter (e.g., the L=8; 10; 15 m beams in Fig. 20),
axial tension leads to an ultimate moment increase, which grows with β and stems
from the improved lateral-torsional buckling resistance.
(ii) The comparison between the Mu /Mpl vs. β curves concerning the (ii1) L=8; 10; 15 m and
(ii2) L=3.5; 5 m beams show different trends, even if all these curves have positive slopes
throughout the whole β range considered. While in the former Mu /Mpl grows with β at
an always increasing rate (upward curvature), which becomes percentage-wise more
relevant as L increases, the latter exhibit points of inflexion, i.e., the curvature changes
from upward to downward at a given β value that seems to increase with L. These
different trends reflect the contradicting influence of axial tension on the lateral-
torsional buckling and cross-section resistances: the latter becomes progressively more
relevant as β increases and L decreases. This assertion is fully confirmed by looking at
the Mu /Mpl vs. β curve concerning the L=2 m beam, which exhibits very little growth
and ends up merging with their L=0.5; 1.0 m beam counterparts for β=2.0 − it would
start descending for larger β values, whenever collapse would start being governed
by plasticity in the beam mid-span region.
(iii) Naturally, the Mu /Mpl percentage growth with β is considerably larger for the longer
(more slender) beams − e.g., for L=15 m and β=2.0, Mu /Mpl increases by almost 85%.
(for L=3.5 m this same increase is just about 27%).
(iv) Concerning the influence of the bending moment diagram shape on the axial tension
benefit, shown in Fig. 21 for the L=15 m S355 steel HEB300 beams, the first important
observation is that only the ψ=1 and ψ=0.5 (marginally) curves (i.e., those leading
to more relevant lateral-torsional buckling effects) are not limited by the descending
curve associated the mid-span cross-section full yielding up to β=2.0 − indeed, the ψ=0
and ψ= − 0.5 curves merge into this curve at lower (decreasing) β values and following
an “almost horizontal” segment. Finally the ψ= − 1 curve decreases monotonically
with ψ, thus meaning that the beam collapse is always governed by the mid-span
cross-section resistance.
(v) Quantitatively speaking, the highest Mu percentage increases due to the presence of axial
tension occur for the beams acted by the ψ=0.5 bending moment diagram − they
are slightly larger than their ψ=0 and ψ=1 diagram counterparts (in this order).
Next, Fig. 22 shows how Mu /Mpl varies with the beam slenderness λLT=(Mpl,Rk /Mcr)0.5,
calculated taking into account the axial tension, for various combination of beam length,
cross-section shape and steel grade. This figure clearly shows that the net effect of the
presence of an increasing axial tension is to move the Mu /Mpl vs. λLT “beam points” (i) to the
left (lateral-torsional slenderness decrease) and (ii) upwards (ultimate moment increase), thus
reflecting the double influence of Nt. Moreover, it can also be observed in this figure that the
whole set of points, corresponding to various beams and β values (including β=0), remain
nicely “aligned” along a “design-like” curve. The design approach for beams subjected to
tension that is proposed in the next section takes advantage of this feature. Moreover, this
design approach it will be validated against the extensive ultimate moment data bank obtained
by means of the parametric study that was just (partially) reported.
684
Figure 22: Variation of Mu /Mpl with the beam slenderness λLT
7
Although no investigation was carried on this matter, the authors believe that such design guidance is also
missing in the vast majority of the current steel structures codes.
8
Class 1 or Class 2, according to the EC3 nomenclature.
9
Method applicable only to hot-rolled and “equivalent welded” beams − any other beam must designed by means
of the so-called “general method”, which is more conservative.
685
(iii) Further modify/increase the reduction factor obtained in the previous step, by
means of the relation χLT.mod =χLT /f, where f ≤ 1.0 is a parameter that depends on the
bending moment diagram and beam slenderness λLT −it supposedly reflects the influence
of the spread of plasticity taking place prior to the beam collapse.
(iv) Evaluate the beam bending resistance against from lateral-torsional buckling failure,
which is termed Mb,Rd and given by Mb,Rd=χLT.mod × Mpl,Rk.
The proposed design approach consists of merely incorporating the axial tension beneficial
effects into the above procedure. This is done exclusively through the value of the critical
buckling moment used to determine the beam slenderness, while keeping all the remaining
steps unchanged − in particular, Mpl,Rk still remains the cross-section pure bending resistance
(i.e., does not account for the presence of axial tension). In other words, Mcr≡Mcr (0) is
replaced by Mcr (Nt,Ed), where Nt,Ed is the acting axial tension, which implies a λLT decrease
and, therefore, also larger χLT.mod and Mb,Rd values. It is worth noting that the calculation of
Mcr (Nt,Ed) must be done by means of a numerical beam buckling analysis (e.g., using beam
finite elements) − in the future, the authors plan to develop analytical expressions and/or other
design aids that will render the performance of this task easier and more straightforward.
Before comparing the obtained numerical and estimated ultimate moments, it should be
pointed out that this comparison concerns exclusively the beams whose collapse does not
correspond to exhausting the beam mid-span cross-section resistance, i.e., beams failing in
lateral-torsional modes occurring prior to the attainment of MN,Rk − indeed, the proposed
design approach only concern the latter. Figs. 23 to 25 concern the various columns analyzed
under bending moment diagrams defined by ψ=1, ψ=0 and ψ= − 1, respectively, and make
it possible to compare the numerical ultimate moments with their predictions provided by the
proposed design approach − use of the EC3 design curve b with a lateral-torsional slenderness
modified through the inclusion of the critical buckling dependence on Nt. The observation of
these three figures prompts the following remarks:
(i) First of all, it is worth noting that the length of the EC3 design curve b horizontal
plateau depends on the bending moment diagram acting on the beam − indeed, this
plateau length increases from 0.4 (ψ=1) to 0.70 (ψ=0) and 0.80 (ψ= − 1).
(ii) Then, it is impossible not to notice the remarkable closeness between the numerical
ultimate moments and their predictions provided by the proposed design approach.
Indeed, in the three figures the numerical values are virtually aligned on top the design
686
Figure 23: Comparison between the Mu /Mpl,Rk (numerical) and Mb,Rd /Mpl,Rk (proposed design approach) for ψ=1
Figure 34: Comparison between the Mu /Mpl,Rk (numerical) and Mb,Rd /Mpl,Rk (proposed design approach) for ψ=0
687
Figure 25: Comparison between the Mu /Mpl,Rk (numerical) and Mb,Rd /Mpl,Rk (proposed design approach) for ψ=− 1
curve − it is only possible to detect a very slight underestimation in the high slenderness
range (it grows with λLT), particularly for ψ=1 (heavier lateral-torsional buckling effects).
(iii) However, it is equally impossible not to notice the few striking exceptions to the very
general rule put forward in the previous item. They all concern the same beam: an
L=25 m S460 steel HEB500 beam. The explanation for these discrepancies lies in the
fact that, for most loadings (i.e., whenever the mid-span cross-section resistance does not
govern), this beam collapses at extremely high deformation levels (e.g., torsional
rotations above 90°), associated to ultimate moments that are clearly underestimated by
the design curve, particularly for ψ=0. If the ultimate moments were linked to
“acceptable deformation levels” (instead of actual equilibrium path limit points) their
values would drop and end up much closer to the design curve. For instance, if the
(perfectly logical) torsional rotation limit of 15° was adopted as a beam ultimate limit
state, all the Mu /Mpl,Rk values associated with the beam under consideration would be
no more than about 12% above the EC3 design curve b (value for ψ =0 and β=1).
(iv) Table 2 provides the averages, standard deviations and maximum/minimum values of
the ratio RM=Mu /Mu.est corresponding to Figs. 23-25 and for the various axial tension
Table 2: Averages, standard deviations and maximum/minimum value of the ratio RM
Average St. Dev. Max Min
β =0 1.08 0.06 1.17 0.98
β =0.5 1.04 0.09 1.11 0.97
β =0.75 1.07 0.10 1.16 0.94
β =1 1.05 0.12 1.09 0.95
β =1.5 1.02 0.08 1.08 0.96
β =2 1.02 0.04 1.08 0.93
688
levels (excluding the 13 L=25 m S460 steel HEB500 beam ultimate moments). These
indicators reflect the excellent quality of the ultimate moment estimates − indeed, the
overwhelming majority of them are safe and extremely accurate. It is still worth noticing
that the least accurate estimations (higher average and standard deviation) concern β=1.
In view of what was mentioned above, it seems fair to conclude that the proposed design
approach for beams subjected to torsion provides excellent estimates of all the numerical
ultimate moments obtained in this work (associated with lateral-torsional collapse modes)
and, therefore, can be considered as a very promising candidate for inclusion in a future
version of Eurocode 3 − of course, additional parametric and reliability studies are needed
before this goal can be achieved. The only foreseeable hurdle for designers is the lack of
an easy and user-friendly way to calculate critical buckling moment in the presence of axial
tension − as mentioned earlier, the authors plan to work on the removal of this hurdle through
the development of analytical expressions and/or other design aids to calculate Mcr (Nt,Ed).
Finally, Table 4 provides the averages, standard deviations and maximum/minimum values
of the percentage ultimate moment increases (∆Mb,Rd) due to axial tension corresponding to
Figs. 23-25 and excluding again the 13 L=25 m S460 steel HEB500 beam ultimate moments.
It is observed that all the above axial tension benefit indicators increase with β, with the sole
exception of the minimum value − it remains constant and very small, because it always
corresponds to a slenderness located very close to the end of the design curve plateau.
Table 3: Ultimate moment predictions for the L=8.0 m S355 steel IPE500 beam under uniform bending
β λLT χLT.mod Mb.Rd [kNm] ∆Mb.Rd [kNm]
0 1.683 0.357 278.1 −
0.5 1.575 0.396 308.9 30.8%
0.75 1.509 0.423 329.9 51.8%
1 1.464 0.443 345.1 67%
1.5 1.350 0.498 387.8 109.7%
2 1.231 0.562 437.8 159.7%
689
Figure 26: Pictorial representation of the ultimate moment predictions − L=8.0 m S355 steel IPE500 beam (ψ=1)
Conclusion
This paper reported the results of an analytical, numerical and experimental investigation on
the lateral-torsional stability, failure and design of hot-rolled steel I-section beams with fork-
type end supports and acted by simple transverse loadings (mostly applied end moments) and
various axial tension values. Initially, the derivation and validation of an analytical expression
providing critical buckling moments of uniformly bent beams subjected to tension was
presented. Then, this analytical finding was followed by a numerical study on the beneficial
influence of axial tension on beams under non-uniform bending, namely caused by unequal
applied end moments or a uniformly distributed load − several beam finite element results
were presented and discussed in some detail. Next, the paper addressed the performance of
two experimental tests, carried out at the University of Fribourg and aimed at determining the
behavior and ultimate strength of a narrow and a wide flange beams subjected to eccentric
axial tension. The experimental set-up and procedure were described, and the main results
obtained (initial imperfection and residual stress measurements, and beam equilibrium paths)
were displayed and briefly commented. These results were also used to develop and validate
FINELG beam and shell finite element models that were subsequently employed to perform
an extensive parametric study that (i) involved more than 2000 numerical simulations,
concerning beams with various cross-section shapes, lengths, yield stresses, acting bending
moment diagrams and axial tension levels, and (ii) was carried to gather a fairly large ultimate
strength/moment data bank. Finally, these data were then used to assess the merits of a design
approach proposed for beams subjected to tension and collapsing in lateral-torsional modes
− this design approach consists of slightly modifying the current procedure prescribed in
Eurocode 3 to design beams against lateral-torsional failures (through the incorporation of the
axial tension influence on the critical buckling moment that is used to evaluate the beam
slenderness). The ultimate moment estimates provided by proposed design approach were
690
shown to correlate extremely well with the values obtained from the numerical simulations,
thus making it a very promising candidate for inclusion in a future version of Eurocode 3.
Lastly, the paper closed with a quick assessment of the beneficial influence of axial tension
on the ultimate strength/moment of beams failing in lateral-torsional modes − as expected, it
was found that this influence can be quite significant and, thus, neglecting it (as is currently
done in practically all steel structures codes) will certainly lead to over-conservative designs.
References
Boissonnade N. Greiner R. Jaspart JP. Lindner J (2006). Rules for Member Stability in EN 1993-1-1: Background
Documentation and Design Guidelines (ECSS Technical Committee 8 − Stability). ECCS Publication Nº 119.
Boissonnade N. Somja H (2012). “Influence of imperfections in FEM modeling of lateral-torsional buckling”.
USB Key Drive Proceedings of SSRC Annual Stability Conference. (Grapevine. 17-21/4).
CEN (Comité Européen de Normalisation) (1992). Eurocode 3: Design of Steel Structures − Part 1.1: General
Rules and Rules for Buildings (ENV 1993-1-1). Brussels.
CEN (Comité Européen de Normalisation) (2005). Eurocode 3: Design of Steel Structures − Part 1.1: General
Rules and Rules for Buildings (EN 1993-1-1). Brussels.
Chen W-F. Atsuta T (1977). Theory of Beam-Columns − Space Behavior and Design (vol. 2). McGraw-
Hill. New York.
Culver C (1966). “Exact solution of the biaxial bending equations”. Journal of the Structural Division (ASCE).
92(2). 63-83.
FINELG (Non-Linear Finite Element Analysis Program) (2012). User’s Manual (vrs. 9.3). ArGEnCo Department.
University of Liège. and Greisch Info S.A.. Liège.
Simulia Inc. (2008). Abaqus Standard (vrs. 6.7-5).
Tomás J (2013). I-Section Steel Beams under Tension: Lateral-Torsional Buckling. Ultimate Strength and Design.
M.A.Sc. Thesis in Civil Engineering (Structures). Instituto Superior Técnico. Technical University of Lisbon.
Trahair NS (1993). Flexural Torsional Buckling of Structures. E&FN Spon (Chapman & Hall). London.
Ville de Goyet V (1989). L’Analyse Statique Non Linéaire par la Méthode des Élements Finis des Structures
Spatiales Formées de Poutres à Section Symétrique. Ph.D. Thesis in Applied Science. University of Liège.
(French)
691
ANNEX A
Table A1: Numerical and design results concerning the L=15 m S355 steel IPE300 beams
ψ =1
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm] Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
0 30.405 223.2 223.181 32.968 2.709 0.155 34.7 0.95
ψ =0.5
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
0 39.960 223.2 223.181 44.456 2.363 0.199 44.4 1.00
ψ =0
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
0 54.493 223.2 223.181 49.911 2.024 0.262 58.4 0.85
ψ = − 0.5
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
ψ =−1
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
692
Table A2: Numerical and design results concerning the L=5 m S460 steel IPE500 beams
ψ =1
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
0 552.7 1009.7 1009.7 464.0 1.352 0.497 501.6 0.93
ψ =0.5
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
0 729.6 1009.7 1009.7 607.0 1.176 0.625 630.7 0.96
ψ =0
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
693
Table A3: Numerical and design results concerning the L=10 m S355 steel HEB300 beams
ψ =1
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
694
Table A4: Numerical and design results concerning the L=25 m S355 steel HEB500 beams
ψ =1
β [m-1] Mcr [kNm] Mpl.Rk [kNm] MN.Rk [kNm Mu [kNm] λLT χLT.mod Mb,Rd [kNm] RM
695